In the case of a cluster configuration having a plurality of servers and a storage server, a server stops in accordance with a load of entire service to reduce the power consumption, thereby making it possible to realize a reduction in running cost. As a realization method for stopping a server group in accordance with a load, a method in which a power management server manages a power status of each server through overconcentration has been proposed (e.g., see Patent Document 1). The power management server monitors load information of the servers and issues a command to stop a power supply of a server having a low load.
Methods for starting up the power supply of the server in such a manner include two methods.
A first method is a method in which the power management server always monitors the load information of the servers and starts up a stopping server when a load of the server group increases. A second method is a method in which a start-up time is set in advance when a server stops and the server starts up autonomously at that time.
It is to be noted that start-up of a machine at a scheduled time includes, for example, start-up using an Advanced Configuration and Power Interface Wakeup function or the like.